Read / Write Image

Accelerated Computing CUDA CUDA Programming and Performance
1

Could anyone point me to a client / kernel example of the following? :

  1. Write to an image_2d in a kernel
  2. Read from the image buffer and write, in the client (c/c++), to an image format such as bmp or jpg.

I have written a kernel to do this but I can’t seem to read from the buffer in the client.

2

To read data from an OpenCL image you can use this function: clEnqueueReadImage() - see documentation for more details
You can then store the data into a bmp/jpg/whatever image but that has nothing to do with OpenCL … to do that , you can use some image libraries like OpenIL or others

3

Thanks for the reply.

My question is specifically directed at what to do once you get the buffer back. For example i have written to a 1024x1024 image in my kernel and read it into an array of bytes in my c++ code.

Is this the correct way of doing it? If I have a 32 bit unsigned int rgba pixel formatted image at 1024x1024 do I just read the image into a buffer that is 4 x 1024 x 1024 and then use that as data to pack into my bitmap?

4

To write into an image in a kernel you have to use the function write_imageui.

Here is an example on how you can copy an image2d into another one using a kernel. (Note that that for this purpose it is better to use the clEnqueueCopyImage() into the “client” API.

__kernel void CopyImage(__read_only image2d_t imageIn,__write_only image2d_t imageOut)

{

  const sampler_t sampler=CLK_NORMALIZED_COORDS_FALSE|CLK_ADDRESS_CLAMP|CLK_FI

LTER_NEAREST;

  int gid0 = get_global_id(0);

  int gid1 = get_global_id(1);

  uint4 pixel;

  pixel=read_imageui(imageIn,sampler,(int2)(gid0,gid1));

  write_imageui (imageOut,(int2)(gid0,gid1),pixel);

}